FIREWORK ALGORITHM DYNAMICS SIMULATED AND ANALYZED...
Transcript of FIREWORK ALGORITHM DYNAMICS SIMULATED AND ANALYZED...
FIREWORK ALGORITHM DYNAMICS SIMULATED AND ANALYZED
WITH THE AID OF COMPLEX NETWORK
Tomas Kadavy
Michal Pluhacek
Adam Viktorin
Roman Senkerik
Tomas Bata University in Zlin, Faculty of Applied Informatics
Nam T.G. Masaryka 5555, 760 01 Zlin, Czech Republic
{kadavy, pluhacek, aviktorin, senkerik }@fai.utb.cz
KEYWORDS
Firework Algorithm, FWA, Complex Network, Surface
Analysis
ABSTRACT
In this paper, we are presenting a visualization of
Firework Algorithm (FWA) inner dynamics as an
evolving complex network. Recent research in
unconventional controlling and simulation of
metaheuristic dynamics shows that this kind of
visualization technique has been utilized only for
algorithms with some social communication or behavior
leading to sharing information across the population. Our
simulation experiment presents the original approach for
analyzing the complex dynamics of an algorithm based
mostly on random/local search engines. Provided
analysis suggests that the built network can be used for
identification of test function surfaces types.
INTRODUCTION
The Firework Algorithm (FWA) is an algorithm for
numerical optimization, which has been introduced in
2010 by authors Tan and Zhu (Tan, Zhu 2010). The
algorithm is based on fireworks explosions in the sky.
This algorithm has similar characteristics like some
scatter search algorithms (Laguna, Marti 2003) or tabu
search algorithm (Glover 1986). This FWA can be
described as non-bio-inspired algorithm like a water drop
algorithm (Shah-Hosseini 2009), brain storm
optimization (Shi 2011) or magnetic optimization
algorithms (Tayarani, Akbarzadeh-T 2008).
In this paper, the inner complex dynamic of FWA is
transformed into the evolving complex network (Barrat
et al. 2008). The population is visualised as an evolving
complex network that usually exhibits non-trivial
features – e.g. degree distribution, clustering, centralities
and in between. These features offer a clear description
of the population under evaluation and can be utilised for
adaptive population as well as parameter control during
the metaheuristic run. Such a complex networks can be
more detailed analyzed using dedicated techniques (Otte
et al. 2002; Kudelka et al. 2015). Recent research shows,
that those analyses have been made for Swarm
Intelligence (SI) algorithms (Pluhacek et al. 2016a;
Pluhacek et al. 2016b; Senkerik et al. 2016a; Senkerik et
al. 2016b), which exhibits some social behavior or
communication across particles (individual solutions).
Our simulation experiment presents the original approach
for analyzing the complex dynamics of an algorithm
based not on social behavior, but mostly on random/local
search engines. The research tasks can be summarized as
follows:
Could the similar behavior or communication be observed using the complex network approach also for FWA?
Can the network features be used to observe and identify the surface of tested function?
The paper is structured as follows. The FWA is described
in details within the next section. The complex network
design and used test functions follow afterwards. Last
two sections discuss the experiment setting and results.
FIREWORK ALGORITHM
The FWA is an algorithm that is inspired by fireworks
explosion in a night sky. This algorithm is initialized with
a random population of fireworks X. The xi firework
position is represented as coordinates in n-dimensional
space of solutions. These coordinates are parameters of
the optimized problem. The number of the fireworks is
defined by parameter NP; this parameter is set by the
user. Moreover, the user defines the parameters like
number of iterations of the algorithm (terminal
condition), Gaussian mutation �̂�, number of sparks m,
parameters a and b and constant �̂�. This algorithm
consists of four parts: explosion operator, mutation
operator, mapping rule and selection strategy. These
parts and adjustable parameters are more explained in
next sections.
The realization of FWA is as follows:
1. Randomly generate NP fireworks in the n-
dimensional search space.
2. Obtain fitness values of these generated
fireworks by fitness function.
3. Calculate the number of generated sparks and
their amplitude for each firework by explosion
operator.
4. Use Gaussian mutation to generate new random
sparks by mutation operator.
5. Apply mapping rule to all generated sparks.
Proceedings 31st European Conference on Modelling and Simulation ©ECMS Zita Zoltay Paprika, Péter Horák, Kata Váradi, Péter Tamás Zwierczyk, Ágnes Vidovics-Dancs, János Péter Rádics (Editors) ISBN: 978-0-9932440-4-9/ ISBN: 978-0-9932440-5-6 (CD)
6. Calculate fitness values of sparks and by
applying selection strategy pick the selected
sparks as new fireworks.
7. If the terminal conditions are met, stop the
algorithm. Otherwise, continue the iteration
process from 3.
There can be more or different terminal conditions
defined by the user. For example, a number of fitness
evaluation (FE) instead of a number of iterations of the
algorithm.
Initialization
The initial NP, the number of fireworks, fireworks X are
randomly generated with uniform distribution from the
range which is specified for the problem by lower and
upper bounds defined by the optimized problem with
dimensionality dim.
In the initialization phase, the adjustable parameters
mentioned before has to be defined as well.
Explosion Operator
The number of sparks generated from each firework is
determined by the firework fitness value. The firework
with better fitness value produces more sparks (the lower
cost function f(x), the better fitness value). This number
of sparks is calculated by explosion strength in (1).
𝑆𝑖 = 𝑚 ∙𝑌𝑚𝑎𝑥−𝑓(𝑥𝑖)+𝜀
∑ (𝑌𝑚𝑎𝑥−𝑓(𝑥𝑖))𝑁𝑃𝑖=1 +𝜀 (1)
where Si is the number of sparks for firework i, m is the
total number of sparks defined by the user. Ymax means
the fitness value of the worst individual (firework).
Function f(xi) is the fitness value for the individual
firework i. The last parameter ε is used to prevent the
denominator from becoming zero and it should be the
smallest possible number.
There is also a limitation of the number of generated
spark defined as (2).
𝑠�̂� = {
𝑟𝑜𝑢𝑛𝑑(𝑎 ∙ 𝑚), 𝑖𝑓 𝑠𝑖 < 𝑎 ∙ 𝑚
𝑟𝑜𝑢𝑛𝑑(𝑏 ∙ 𝑚), 𝑖𝑓 𝑠𝑖 > 𝑏 ∙ 𝑚
𝑟𝑜𝑢𝑛𝑑(𝑠𝑖), 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
(2)
where a and b are constants defined by the user (these
constants has to be a<b<1), 𝑠�̂� is the limitation of the
number of sparks and round() is the rounding function.
The amplitude for generated sparks is then calculated by
explosion amplitude in (3). Like the previous, explosion
strength, the amplitude of explosion is defined by
firework fitness function. The better fitness value is, the
smaller is the amplitude of explosion and vice versa.
𝐴𝑖 = �̂� ∙𝑓(𝑥𝑖)−𝑌𝑚𝑖𝑛+𝜀
∑ [𝑓(𝑥𝑖−𝑌𝑚𝑖𝑛)]𝑁𝑃𝑖=1 +𝜀
(3)
where Ai is the amplitude of i firework. �̂� is a constant
defined by the user and means the sum of all amplitudes.
Ymin means the fitness value of the best firework.
The new sparks are generated in randomly chosen
dimensions z and the position is calculated in (4).
�̂�𝑗𝑘 = 𝑥𝑖
𝑘 + 𝑈(−𝐴𝑖 , 𝐴𝑖) (4)
where �̂�𝑗𝑘 is spark j in dimension k (𝑘 ∈ 𝒛) generated from
firework xi. U is a random number from a uniform
distribution in the range of the explosion amplitude of i
firework.
Mutation Operator
To maintain the diversity of the population, some
mutation operator is needed. For FWA, the Gaussian
mutation is used. The sparks are generated as follows:
1. Choose random firework i.
2. Compute new spark using formula (5).
3. If the number of generated spark by Gaussian
mutation reaches the value �̂�, stop generating
next sparks
�̂�𝑗𝑘 = 𝑥𝑖
𝑘 ∙ 𝑁(1, 1) (5)
where �̂�𝑗𝑘 is spark j in dimension k (𝑘 ∈ 𝒛) generated from
firework xi. Vector z are randomly chosen dimensions
like in section Explosion Operator. N is a random number
from normal (Gaussian) distribution with mean 1 and
variance 1.
Mapping Rule
This rule ensures, that all previously generated sparks are
in feasible space. If any spark lies outside of the available
search space, its mapped back to allowed space. This
mapping rule defined as (6).
�̂�𝑖𝑘 = 𝐵𝐿
𝑘 + �̂�𝑖𝑘𝑚𝑜𝑑(𝐵𝑈
𝑘 − 𝐵𝐿𝑘) (6)
where �̂�𝑖𝑘 is i particle in k dimension, 𝐵𝐿
𝑘 and 𝐵𝑈𝑘 are lower
and upper boundaries of the available search space in k
dimension. The mod represents modular operation.
Selection Strategy
Some of the generated sparks need to be selected and
passed into the new iteration. These selected sparks will
become new fireworks. For this selection, the distance-
based strategy is used to maintain the diversity of the
population. The spark that is farther from the others has
the greater chance to be selected than those sparks near
the other sparks. The first chosen spark is always the one
with the best fitness value. Others (NP-1) individuals are
chosen by roulette method. The possibility of choosing
the spark into next iteration is calculated in (7).
𝑝𝑖 =𝑅𝑖
∑ 𝑅𝑗𝐾𝑗=1
(7)
where pi is the possibility of the i spark, Ri is the sum of
distances of the i spark, K is the number of all generated
sparks. The Euclidean distance is used to compute the Ri
in formula (8).
𝑅𝑖 = ∑ 𝑑(�̂�𝑖 , �̂�𝑗)𝐾𝑗=1 = ∑ ‖�̂�𝑖 − �̂�𝑗‖𝐾
𝑗=1 (8)
where K is the number of all sparks, �̂�𝑖is the spark for
which the Ri is computed and �̂�𝑗 are others sparks where
𝑗 ∈ 𝐾.
The whole FWA is depicted in the pseudo-code below.
Algorithm pseudo-code 1: FWA
1. Randomly initialize NP fireworks
2. while terminal condition not met
3. count fireworks fitness values
4. for i = 1 to NP do
5. calculate Si
6. calculate Ai
7. generate sparks of i firework
8. end
9. for j = 1 to �̂� do
10. Gaussian mutation
11. end
12. selection strategy for new
fireworks
13. end
NETWORK DESIGN
The network is created as a history of contributions. In
each iteration, there are NP fireworks. These fireworks
create K sparks. Some of these sparks are transferred into
a new iteration as new fireworks. Fireworks are then
represented as the nodes in the network. These nodes are
labelled 1…NP for each iteration. The nodes (fireworks)
are sorted by their fitness values before labelling so that
the best node (smallest fitness value) gets number 1 and
the worst node gets number NP. The edge between nodes
represents spark that creates a new firework in next
iteration. The initial node of the edge represents the
firework from which the spark is created. The terminal
node is the firework in the next iteration created by the
spark. With that rule, the initial node from t iteration can
have from 0 to NP edges and terminal node can only have
one edge as input.
An example of the network with five fireworks in four
iterations is shown in Figure 1. Blue edges indicate the
spark with the best fitness function value. The blue edge
direction can only be towards the node number one. The
first iteration is on left side of the figure, and the last
iteration is on the right side. From the first iteration, four
sparks create new fireworks in the second iteration and
one of them contributes to improving the solution.
Figure 1: The example of FWA network.
TEST FUNCTIONS
For the simulation experiment, a set of 5 classic functions
were selected. The set consists of unimodal and
multimodal functions:
Sphere function (f1) (9),
Rosenbrock function (f2) (10),
Rastrigin function (f3) (11),
Schwefel function (f4) (12),
Egg holder function (f5) (13).
𝑓(𝑥)1 = ∑ 𝑥𝑖2𝑑𝑖𝑚
𝑖=1 (9)
𝑓(𝑥)2 = ∑ [100 ∙ (𝑥𝑖+1 − 𝑥𝑖2)
2+ (1 − 𝑥𝑖)2]
𝑑𝑖𝑚−1
𝑖=1 (10)
𝑓(𝑥)3 = 10 ∙ 𝑑𝑖𝑚 + ∑ [𝑥𝑖2 − 10 ∙ 𝑐𝑜𝑠(2𝜋𝑥𝑖)]
𝑑𝑖𝑚
𝑖=1 (11)
𝑓(𝑥)4 = ∑ [−𝑥𝑖 ∙ 𝑠𝑖𝑛(|𝑥𝑖|0.5)]𝑑𝑖𝑚
𝑖=1 (12)
𝑓(𝑥)5 = ∑ [−(𝑥𝑖+1 + 47) ∙ 𝑠𝑖𝑛 (√|𝑥𝑖+1 +𝑥𝑖
2+ 47|) − 𝑥𝑖
𝑑𝑖𝑚−1
𝑖=1
∙ 𝑠𝑖𝑛 (√|𝑥𝑖 − (𝑥𝑖+1 + 47)|)] (13)
EXPERIMENT SETTING
The experiments were performed for test functions
dimensions 2, 10 and 30. The number of iterations was
set to 45. The control FVA parameters were set
accordingly to (Tan, Zhu 2010). The number of
fireworks, population size (NP), was set to 5 for all
dimensions. The number of sparks (m) was set as 50.
Parameters a and b were set as 0.8 and 0.04. Other
constant settings were following: �̂� = 40 and �̂� = 5.
The basic logical assumption was that the longest path of
steady improvement in the network (i.e. the path between
1 1
2
4
3
4
5
2
3
5
1
2
3
4
5
1
2
3
4
5
nodes labeled 1 and joined with blue edges) would be
observable mostly for the unimodal function (e.g. f1).
RESULTS
The results for the aforementioned longest paths of the
stable improvement are given in Table 1.
Table 1: Longest paths of steady improvement in the network.
Function Dimension
2 10 30
f1 10 8 15
f2 20 10 8
f3 11 6 7
f4 9 8 7
f5 5 5 15
Results depicted in Table 1 confirm the anticipated
logical assumption made in the previous section. For the
unimodal functions (f1 and f2), the observed path is quite
longer compared to the results for the multimodal
functions. The differences are decreasing with the higher
dimension setting. These trends are graphically
confirmed also in Figures 2 - 7. Nonetheless, the data
shown in Figure 8 as well as in Table 1, indicate an
exception to the primary logical assumption. Detailed
analysis reveals that the path of steady improvement is
present at the beginning of the captured evolved network
(optimization process) and this may be caused by
premature stagnation in the local optimum of tested
function. For unimodal functions, the path of steady
improvement seems to be present more often at the end
of the recorded optimization process.
CONCLUSION
In this paper, the possibility of simulation and simple
analysis of complex network evolvement for firework
algorithm inner dynamics is present. Our novel approach
was tested on the set of five simple classical benchmark
functions.
The preliminary results lend weight to the argument that
the ability of a network to identify the surface type of
optimised function seems to be present. Nevertheless,
more and detailed in-depth study is required to be
performed in this field.
Another phenomenon has been discovered. The network
seems to have a lack of any other usable information. The
results of this simple simulation study will be further
used in future research to suggest possible improvements
to building complex networks for the family of
algorithms based on the local/random search techniques
without accessible direct social/communication
interactions.
Figure 2: Network of f2 for dim 2.
Figure 3: Network of f5 for dim 2.
Figure 4: Network of f2 for dim 10.
3
1
4
2
3
2
4
1
55
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1 1
2
3 3
4
5
2
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
2
1
5
2
33
4
1
5
4
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
Figure 5: Network of f5 for dim 10.
Figure 6: Network of f1 for dim 30.
Figure 7: Network of f4 for dim 30.
Figure 8: Network of f5 for dim 30.
ACKNOWLEDGEMENT
This work was supported by Grant Agency of the Czech
Republic – GACR P103/15/06700S, further by the
Ministry of Education, Youth and Sports of the Czech
Republic within the National Sustainability Programme
Project no. LO1303 (MSMT-7778/2014). Also by the
European Regional Development Fund under the Project
CEBIA-Tech no. CZ.1.05/2.1.00/03.0089 and by Internal
Grant Agency of Tomas Bata University under the
Projects no. IGA/CebiaTech/2017/004.
REFERENCES
Tan Y., Zhu Y. (2010) Fireworks Algorithm for Optimization. In: Tan Y., Shi Y., Tan K.C. (eds) Advances in Swarm Intelligence. ICSI 2010. Lecture Notes in Computer Science, vol 6145. Springer, Berlin, Heidelberg
Laguna M., Marti R. Scatter search: methodology and implementations in C. Boston: Kluwer Academic Publishers, c2003. ISBN 9781402073762.
Glover F., Future paths for integer programming and links to artificial intelligence, Computers & Operations Research, Volume 13, Issue 5, 1986, Pages 533-549
Shah-Hosseini H., the intelligent water drops algorithm: a nature-inspired swarm-based optimization algorithm. Int. J. Bio-Inspir. Comput. 1(1), 71-79 (2009)
Shi Y., Brain storm optimization algorithm, in Advances in Swarm intelligence (Springer, Berlin, 2011), pp. 303-309
Tayarani N.M.H., Akbarzadeh-T M.R., Magnetic optimization algorithms a new synthesis, in 2008 IEEE World Congress on Compunational Intelligence Evolutionary Computation (CEC) (IEEE, 2008), pp. 2659-2664
Barrat, A., Barthelemy M., Vespignani A. Dynamical processes on complex networks. New York: Cambridge University Press, 2008. ISBN 9780521879507.
Otte, Evelien; Rousseau, Ronald (2002). "Social network analysis: a powerful strategy, also for the information sciences". Journal of Information Science. 28 (6): 441–453
1 1
2 2
4
3
4
3
55
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
4
1
2 2
5
3
4
5
1
3
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
2
1
5
2
3 3
4
1
5
4
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
4
1
5
2
3 3
2
4
5
1 1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
Kudĕlka, M., Zehnalová, Š., Horák, Z., Krömer, P., & Snášel, V. (2015). Local dependency in networks. International Journal of Applied Mathematics and Computer Science, 25(2), 281-293.
Pluhacek, M., Janostik, J., Senkerik, R., & Zelinka, I. (2016a). Converting PSO dynamics into complex network-Initial study. In T. Simos, & C. Tsitouras (Eds.), AIP Conference Proceedings (Vol. 1738, No. 1, p. 120021). AIP Publishing.
Pluhacek, M., Senkerik, R., Janostik, J., Viktorin, A., & Zelinka, I. (2016b). Study on swarm dynamics converted into complex network. In Proceedings-30th European
Conference on Modelling and Simulation, ECMS 2016. European Council for Modelling and Simulation (ECMS).
Senkerik, R., Viktorin, A., Pluhacek, M., Janostik, J., & Davendra, D. (2016a). On the Influence of Different Randomization and Complex Network Analysis for Differential Evolution. In 2016 IEEE Congress on Evolutionary Computation (CEC) (pp. 3346-3353). IEEE.
Senkerik, R., Viktorin, A., Pluhacek, M., Janostik, J., & Oplatkova, Z. K. (2016b). Study on the Time Development of Complex Network for Metaheuristic. In Artificial Intelligence Perspectives in Intelligent Systems (pp. 525-533). Springer International Publishing.
AUTHOR BIOGRAPHIES
TOMAS KADAVY was born in the Czech Republic,
and went to the Faculty of Applied
Informatics at Tomas Bata University in
Zlín, where he studied Information
Technologies and obtained his MSc degree
in 2016. He is studying his Ph.D. at the
same university and the fields of his studies
are: Artificial intelligence and evolutionary algorithms.
His email address is: [email protected]
MICHAL PLUHACEK was born in the Czech
Republic, and went to the Faculty of
Applied Informatics at Tomas Bata
University in Zlín, where he studied
Information Technologies and obtained his
MSc degree in 2011 and Ph.D. in 2016 with
the dissertation topic: Modern method of
development and modifications of evolutionary
computational techniques. He now works as a researcher
at the same university. His email address is:
ADAM VIKTORIN was born in the Czech Republic,
and went to the Faculty of Applied
Informatics at Tomas Bata University in
Zlín, where he studied Computer and
Communication Systems and obtained his
MSc degree in 2015. He is studying his
Ph.D. at the same university and the fields
of his studies are: Artificial intelligence, data mining and
evolutionary algorithms. His email address is:
ROMAN SENKERIK was born in the Czech Republic,
and went to the Tomas Bata University in
Zlin, where he studied Technical
Cybernetics and obtained his MSc degree in
2004, Ph.D. degree in Technical
Cybernetics in 2008 and Assoc. prof. in
2013 (Informatics). He is now an Assoc.
prof. at the same university (research and courses in:
Evolutionary Computation, Applied Informatics,
Cryptology, Artificial Intelligence, Mathematical
Informatics). His email address is: [email protected]